MMLU

MMLU (Massive Multitask Language Understanding) - a benchmark of multitask accuracy across 57 tasks.

Category

Knowledge

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for MMLU

Model Score Date Verified
GPT-OSS-120B
99.67%
08.08.2025 Verified
GPT-OSS-20B
84.26%
05.08.2025 Verified
DeepSeek-V3.2
92.46%
26.06.2026 Verified
DeepSeek-V4-Flash
95.41%
26.06.2026 Verified
DeepSeek-V4-Pro
100.00%
26.06.2026 Verified
Gemma 3 4B PT
59.60
27.02.2025 Verified
Gemma 3 12B PT
48.85%
27.02.2025 Verified
Gemma 3 27B PT
62.30%
27.02.2025 Verified
GLM-5.3-Flash
93.44%
27.08.2026 Verified